Coactivator OBF-1 Makes Selective Contacts with Both the POU-Specific Domain and the POU Homeodomain and Acts as a Molecular Clamp on DNA

Abstract

The lymphoid-specific transcriptional coactivator OBF-1 (also known as OCA-B or Bob-1) is recruited to octamer site-containing promoters by interacting with Oct-1 or Oct-2 and thereby enhances the transactivation potential of these two Oct factors. For this interaction the POU domain is sufficient. By contrast, OBF-1 does not interact with the POU domains of other POU proteins, such as Oct-4, Oct-6, or Pit-1, even though these factors bind efficiently to the octamer motif. Here we examined the structural requirements for selective interaction between the POU domain and OBF-1. Previous data have shown that formation of a ternary complex among OBF-1, the POU domain, and the DNA is critically dependent on residues within the octamer site. By methylation interference analysis we identified bases that react differently in the presence of OBF-1 compared to the POU domain alone, and using phosphothioate backbone-modified probes in electrophoretic mobility shift assays, we identified several positions influencing ternary complex formation. We then used Oct-1/Pit-1 POU domain chimeras to analyze the selectivity of the interaction between OBF-1 and the POU domain. This analysis indicated that both the POU specific domain (POU_S) and the POU homeodomain (POU_H) contribute to complex formation. Amino acids that are different in the Pit-1 and Oct-1 POU domains and are considered to be solvent accessible based on the Oct-1 POU domain/DNA cocrystal structure were replaced with alanine residues and analyzed for their influence on complex formation. Thereby, we identified residues L6 and E7 in the POU_S and residues K155 and I159 in the POU_H to be critical in vitro and in vivo for selective interaction with OBF-1. Furthermore, in an in vivo assay we could show that OBF-1 is able to functionally recruit two artificially separated halves of the POU domain to the promoter DNA, thereby leading to transactivation. These data allow us to propose a model of the interaction between OBF-1 and the POU domain, whereby OBF-1 acts as a molecular clamp holding together the two moieties of the POU domain and the DNA.

The transcriptional coactivator called OBF-1 (34), OCA-B (19, 20), or Bob-1 (9) is a proline-rich protein that interacts specifically with both octamer-binding factors present in B cells, Oct-1 and Oct-2. It is generally assumed that this interaction is crucial for octamer motif-mediated gene activation in lymphoid cells. OBF-1 expression is highly cell specific and is found in B lymphocytes of all developmental stages (31, 34) as well as transiently in T lymphocytes upon activation (29, 42). Several studies have shown that OBF-1 can activate transcription in vivo or in vitro by being recruited by Oct-1 or Oct-2 to the conserved octamer site of immunoglobulin (Ig) and other promoters (20, 34). While the Ig promoter is well activated by OBF-1 in such assays, the histone H2B promoter is only weakly activated, suggesting that OBF-1 trans-activates in a promoter-specific manner (20, 34). In addition, OBF-1 activates only promoter and not enhancer octamer sites, independently of their orientation (27, 31).

Targeting of the OBF-1 gene in the mouse has shown that, surprisingly, B-cell development and initial Ig gene transcription are unaffected in the absence of this cofactor (14, 24, 30). Yet, in vivo immune responses and germinal center formation are dramatically impaired in these mice, suggesting that critical target genes for these functions must lie downstream of OBF-1. The defect appears to be primarily an intrinsic B-cell defect because T-cell function in these mice is largely normal (30).

Previous studies have shown that OBF-1 forms a ternary complex with the DNA and full-length protein or the POU domain of Oct-1 or Oct-2 (POU1 or POU2) but not with Oct-4, Oct-6, or Pit-1 (9, 20, 34). In addition, OBF-1 appears to influence neither the on rate nor the off rate of the POU domain-DNA interaction and was shown in vitro to interact with the POU domain also in the absence of DNA (27, 31). Interestingly, not every site that binds Oct-1 or Oct-2 allows formation of a complex with OBF-1, and it was found that in particular the base at the fifth position of the octamer motif (ATGCAAAT) is crucial and has to be an A (4, 8). Furthermore, it has been shown that the N terminus of OBF-1 (amino acids 1 to 118) interacts weakly with the octamer site in the absence of the POU domain (4). In vivo coactivation by OBF-1 through different sites correlates well with the complex formation observed in vitro (4, 8). Yet other residues are clearly also important, because there are sites, such as the interleukin 2 distal octamer site or the gonadotropin-releasing hormone enhancer site (5), that have a conserved position +5 but fail to allow ternary complex formation in vitro or to be activated in vivo (reference 8 and our unpublished data).

The POU domain is a bipartite DNA binding protein module (12) that binds selectively to the DNA octamer motif ATGCAAAT and a subset of derivatives. The POU1-DNA cocrystal structure (16) shows two globular domains, the N-terminal POU specific domain (POU_S) and the C-terminal POU homeodomain (POU_H), that are physically connected by the POU linker. The latter is most divergent in sequence and length among the POU domains and is not visible in the cocrystal, indicating an unstructured conformation. By contrast, crystal structures of POU1 and the Pit-1 POU domain in addition to nuclear magnetic resonance data for POU2 show that the POU_S and the POU_H are highly conserved in this protein family at the primary, secondary, and tertiary structure levels. The POU_S and POU_H both have a helix-turn-helix (HTH) motif with the third helix binding to the DNA in the major groove. The HTH in the POU_S is similar to the one found in prokaryotic proteins like, for example, λ-repressor, cro, or 434 (1, 6, 26). The HTH in the POU_H is homologous to the one found in the eukaryotic homeodomain protein family members like, for example, engrailed or antennapedia (15, 25). Therefore, the POU domain can be regarded as a conglomerate of a classical eukaryotic homeodomain and a prokaryotic DNA binding domain (reviewed in references 11 and 38). The POU_S binds to the consensus subsite, ATGC, of ATGCAAAT, while the POU_H binds to the homeo consensus-derived subsite, AAAT. In the POU1-DNA (H2B promoter octamer) cocrystal the POU domain binds as a monomer, with POU_S and POU_H contacting opposite sides of the DNA double helix, whereas in the Pit-1 POU-DNA (binding site derived from the prolactin gene) cocrystal the POU domain binds as a dimer, with POU_S and POU_H touching the same face of the DNA (13). This indicates an extreme flexibility influenced by the sequence of the DNA binding site.

A number of transcriptional coactivators have been shown to interact with the POU domain(s) of Oct-1 and/or Oct-2. The herpes simplex virus protein VP16 selectively interacts with the POU1_H domain through position E22; in POU2_H this position is a divergent alanine which precludes complex formation (18, 28). In addition, VP16 needs the octamer-flanking DNA sequence TAATGArAT (homeodomain binding site underlined) for efficient interaction with POU1 as well as a host cell factor (17, 40).

The RNA polymerase (Pol) II- and Pol III-specific multiprotein transcriptional coactivator SNAPc/PTF interacts, through the largest of its four subunits, SNAP190 (41), with the POU_S position E7 of both the POU1 and POU2 domains (21, 23). SNAPc/PTF itself binds to a conserved proximal sequence element found in close proximity to an octamer site found in many small nuclear RNA (snRNA) promoters. This binding is independent of the interaction with the POU domain, but it is enhanced approximately 8- to 10-fold through cooperative binding with the POU domain (21).

At least two positions of the POU domain are important for complex formation with OBF-1: L53 and N59 (8). However, these positions are conserved throughout POU domains and therefore cannot account for the selectivity with which OBF-1 interacts only with certain POU domains.

Here we analyzed the selective determinants in the POU domain and in the DNA required for complex formation with OBF-1. We identified base and backbone positions of the DNA which influence the efficiency of ternary complex formation when modified, indicating that OBF-1 is in close proximity to these positions. We found, in agreement with these findings, that OBF-1 can be UV cross-linked to the DNA site, together with the POU domain and also by itself weakly. We identified residues within the POU domain crucial for selective interaction with OBF-1 in both the POU_S and the POU_H. Mutation of these critical residues disrupted interaction in vitro and also in an in vivo recruitment assay. Furthermore, with an in vivo assay we showed that OBF-1 is able to functionally recruit two artificially separated moieties of the POU domain to the promoter DNA, thereby leading to transactivation. Our data therefore support a model whereby OBF-1 clamps the POU domain together with the DNA by interacting with both subunits of the POU domain while bridging the DNA within the octamer motif at several positions.

MATERIALS AND METHODS

All recombinant DNA work was done according to standard procedures, and details of the plasmid constructs and methods used are available on request. Forward primers are written in uppercase, and reverse primers are written in lowercase. Single letters in the other style indicate mutations.

Site-directed mutagenesis, in vitro translation, and electrophoretic mobility shift assays (EMSAs).

Plasmids for the Oct-1/Pit-1 POU domain chimeras were kindly supplied by W. Herr (Cold Spring Harbor Laboratory). Coding sequences thereof, with or without mutations, were amplified by PCR followed by sequential in vitro T7 transcription and reticulocyte lysate translation (Promega). The following sequences were used: CTATTTAGGTGACACTATAGAAACAGACACCATGGAGGAGCCCAGTGA (amplification of OXxX chimeras), CTATTTAGGTGACACTATAGAAACAGACACCATGGACTCCCCGGAA (amplification of PXxX chimeras), caggatcctatgggttgattctttttctt (amplification of XXxO chimeras), caggatcctacgttttcacccgtttttc (amplification of XXxP chimeras), caggatcctacgttttcacccgtttttctTtc (XXxP with R155K), caggatcctacgttttGaTccgtttttctctc (XXxP with V159I), caggatcctacgtAttgacccgtttttctctc (XXxP with K160N), caggatcctacgGtttcacccgtttttctctc (XXxP with T161P), caggatcctacgttttGaGccgtttttctTtc (XXxP with V159I-R155K), caggatcctacgGAttcacccgtttttctctc (XXxP with T161P-K160N), and caggatcctacgGAttGaTccgtttttctTtc (XXxP with T161P-K160N-V159I-R155K).

The vectors pBGO-ATG-POU1 and pBGO-ATG-POU6 (34) were used for introduction of mutations with QuickChange (Stratagene) following the instructions of the manufacturer (primers used for site-directed mutagenesis are available on request). All mutations were verified by sequencing. The POU domain of zebra fish pou-2 was amplified by PCR with zfpou2 DNA (10) as a template and the following two primers: GGTGGATCCAGAGGAGACTCTGACTACTGAAG and gccctaggccaaagctagacgtttcccttctg. The amplified POU domain was then cloned into pBGO-ATG to give pBGO-ATG-zfpou2. OBF-1 was expressed from the vector pBGO-ATG-OBF-1/9. For in vitro translation of all pBGO vectors the TNT reticulocyte lysate kit (Promega) was used following the instructions of the manufacturer. Reticulocyte lysate translation reactions included [³⁵S]Met. Equivalent protein expression of the different mutant proteins was verified by sodium dodecyl sulfate-polyacrylamide gel electrophoresis followed by autoradiography and PhosphorImager quantification.

Unless otherwise specified EMSA reactions were performed as previously described (34) with a ³²P-labeled 50-bp SalI DNA fragment containing the octamer site from the intron Ig heavy chain enhancer.

Transfection and luciferase assays.

All cDNAs to be expressed in Drosophila cells were subcloned into pBD1119 (kindly supplied by E. Hafen, University of Zürich) after KpnI-XbaI removal of green fluorescent protein. In this plasmid cDNA expression is under the control of the Drosophila α1-tubulin promoter. Oct-1 POU domains were subcloned via KpnI-XbaI from the plasmid pBGO-ATG-POU1 (wild type [wt]; L6A, E7A, or L6A-E7A) into pBD1119.

The mutation Oct-6-S6L-D7E was introduced into the Oct-6 cDNA by PCR with the following primers: GCGGAATTCGGCATGGCCACCACCGCGCAG (mOct6 cDNA start Met with EcoRI), ggcgaactgctccaggtcttcgaggctgggagcatcctcgtc (mOct6 POU S6L-D7E with BbsI), CCCAGCCTCGAAGACCTGGAGCAGTTCGCCAAG (mOct6 POU S6L-D7E with BbsI), and caggatatcgggtcactgcacagagccgggc (mOct6 cDNA stop with EcoRV).

The constructs for the POU_S and POU_H domains were made by PCR with the following primers: ACTCgaagacTGGAGGAGCCCAGTGACCTTG, tcctggaga aGTCTTCtaacttgggctggagagggacg, TCCAgaagacGTGCCCTGAATTCTCCA GG, and ggtcccaccaGTCTTCtatgggttgattcttttttctttctgg. Amplified fragments were cleaved with BbsI, treated with T4 DNA polymerase, cloned into the vector pBGO-ATG cleaved with NcoI and filled in with T4 DNA polymerase. After verification by sequencing they were subcloned via KpnI-XbaI into pBD1119. All mutations were verified by sequencing.

The reporter plasmids 8xOcta-Luc and 8xOcta/mut-Luc are described elsewhere.

Drosophila Schneider cells (SL2) were transfected as previously described (34). In brief, transfection with the calcium-phosphate coprecipitation was carried out by using 4 μg of each plasmid (reporter and expression vectors) 4 to 6 h after plating of 10⁷ cells/10-cm-diameter dish. In each sample, the total amount of expression vector DNA was kept constant by the addition of pBD1119-empty, and the total amount of DNA was brought to 20 μg with salmon sperm DNA. Luciferase extracts were prepared 48 h after transfection, and equivalent amounts of protein (100 μg) were used for luciferase assays.

Methylation interference assay and DNA backbone modification.

Methylation interference assay was performed as previously described (32) by using probes amplified by PCR from the H2B promoter or from the Ig(κ) light chain promoter. The sequences of these probes were as follows: (top strand only): H2B, GACACAAGACTTCAACTCTTCACCTTATTTGCATAAGCGATTCTATATAAAAGCGCCTTGTCATACCCT; Ig(κ), CCAATCCTAACTGCTTCTTAATAATTTGCATACCCTCACIGCATCGCCTTGGGGACTTCTTTA. Densitometric quantification of the autoradiographs was performed with NIH Image version 1.5.

The phosphothioate-modified probes were synthesized by Ciba Basel or Microsynth, based on the H2B probe used for the methylation interference assay, and had the following sequence: CGCTTATGCAAATAAGGTG, accttatttgcataagcg.

RESULTS

Multiple positions in the DNA influence the formation of a ternary complex with the POU domain and OBF-1.

To examine how the POU domain/OBF-1 complex is assembled on DNA, we performed methylation interference experiments. For this, two different probes containing consensus octamer sites were partially methylated in vitro with dimethyl sulfate and were used for preparative EMSA reactions with the Oct-1 POU domain either alone or in the presence of OBF-1. The various complexes were resolved by gel electrophoresis, and the corresponding DNA was analyzed by chemical cleavage after elution. In the presence of the POU domain a clear footprint was observed over the octamer motif (compare lanes 3 and 8 with lanes 4 and 9 in Fig. 1A and B) as expected; addition of OBF-1 did not enlarge the footprint, nor did it modify the observed pattern outside of the octamer site (Fig. 1A and B, lanes 5 and 10), in agreement with data from chemical assays and recent DNase I footprint (3, 27). Interestingly, on the two probes used, the octamer site from the H2B or the Ig(κ) promoter, the G at position +3 in the sequence ATGCAAAT was found to be reproducibly underrepresented in the binary complex (POU/DNA; compare lanes 8 and 9 in Fig. 1A and B) and even more so in the ternary complex (compare lane 9 with lane 10 in Fig. 1A and B), indicating that methylation of this position interferes with binding of the POU domain and of OBF-1. The densitometric quantification of these primary results presented in the lower part of Fig. 1 shows that position +3 was about 10-fold reduced selectively in the presence of OBF-1. This position of the octamer site is known from the cocrystal studies of Klemm et al. (16) to interact directly with Arg 49 of the Oct-1 POU_S. By contrast, the other positions of the octamer site examined by methylation interference were either not selectively altered in the presence of OBF-1 or were altered only with the H2B probe (e.g., the G at position 4 on the lower strand).

FIG. 1.

Methylation interference assay showing an altered pattern in the presence of OBF-1. End-labeled DNA probes containing the octamer site from the Ig(κ) light chain promoter (A) or from the histone H2B promoter (B) were used to perform a methylation interference assay with in vitro-translated Oct-1 POU domain and OBF-1. After preparative EMSA the DNA present in the different complexes was eluted, subjected to chemical cleavage, and displayed on a sequencing gel. The products of G and G plus A sequencing reactions of the probes were loaded in lanes 1, 2, 6, and 7. Lanes 3 and 8 contain DNA corresponding to the free probe, lanes 4 and 9 contain DNA present in the POU domain complex, and lanes 5 and 10 contain DNA present in the ternary complex. Lanes 1 through 5, top strand; lanes 6 through 10, bottom strand (for the sequences of the probes used see the Materials and Methods section). The position of the octamer site is boxed in the G plus A lane and schematically indicated by the gray box on the side of the gel (panels A and B). For the H2B probe (panel B) the open rectangle represents the TATA box that was also present in the probe used. The lower part of the figure shows a densitometric quantification of the bands corresponding to the guanines and adenines of either strand of the octamer site in the above gels. The relative intensity (ri) of the bands was plotted for the free DNA (white bars) or for the DNA present in a complex with the POU domain alone (hatched bars) or with the POU domain and OBF-1 (black bars). The arrows point to the strong reduction in intensity of the methylated guanine at position +3 in the complex with OBF-1. For simplicity the octamer sequence is written throughout this paper in the orientation 5′-ATGCAAAT-3′, even though for both the Ig(κ) and H2B promoters this corresponds to the sequence of the lower strand.

We next looked at the potential influence of the DNA backbone on ternary complex formation and designed a number of phosphothioate-modified probes which, on the basis of the Oct-1 POU/DNA cocrystal, should not interfere with binding of the Oct-1 POU domain. These probes were assayed by EMSA and compared with the unmodified probe (Fig. 2A). By quantification of the amount of probe present in the different complexes, we assessed the effects of these modifications. As shown in the quantification presented in Fig. 2B, modification of the phosphate groups at the sense strand positions p4 and p5 (lanes 2 and 1 in Fig. 2A) slightly increased ternary complex formation. Likewise, modification of the phosphate groups at the antisense strand positions αp7, αp8, and αp9 (Fig. 2A, lanes 7, 6, and 5) also increased ternary complex formation, while modification of αp6 (Fig. 2A, lane 8) almost completely eliminated binding of OBF-1 without interfering with binding of the POU domain.

FIG. 2.

Modifications of the DNA backbone within the octamer motif with phosphothioate groups influence ternary complex formation. Oligonucleotides were synthesized with specific backbone phosphate groups replaced by phosphothioate groups (positions p3, p4, and p5 on the top strand and positions αp6, αp7, αp8, and αp9 on the bottom strand; see panel E for a diagram and the Materials and Methods section for the complete sequences of the probes). These modified probes were then tested by EMSA. (A) Autoradiograph from a representative EMSA performed with the Oct-1 POU domain (lanes 1 to 8), OBF-1 (lanes 1 to 8), and modified (lanes 1 to 3 and 5 to 8) or unmodified (lane 4) DNA probes. The position of the modification within the octamer sequence (p5, p4, p3, αp9, αp8, αp7, αp6) is indicated on the right of the autoradiograph. The unmodified DNA probe is indicated by wt. All lanes contained equal amounts of OBF-1 and POU domain. (b) Quantification by PhosphorImager of the amounts of probe present in the different complexes on the basis of five independent experiments. Open bars, POU/DNA complex; solid bars, Oct-1 POU domain (lanes 2 to 5, 7 to 10, 12 to 15, and 17 to 20), and unmodified (lanes 1 to 5) or modified (lanes 6 to 20) DNA probes, as indicated. (D) representative EMSA performed with a constant excess of the Oct-1 POU domain (lanes 21 to 40), increasing amounts of OBF-1 (lanes 22 to 25, 27 to 30, 32 to 35, and 37 to 40), and unmodified (lanes 21 to 25) or modified (lanes 26 to 40) DNA probes, as indicated. (E) Sequence of the central part of the probe with circles indicating the modified positions. ⊕, positions which when modified increase ternary complex formation; ⊝, position which when modified reduce ternary complex formation; ○, position which when modified is neutral.

To rule out the possibility that the observed effects might merely reflect an altered interaction between the POU domain and the DNA due to some of the backbone modifications, these experiments were repeated either in the complete absence of OBF-1 (data not shown) or under conditions where the amount of one protein (POU domain or OBF-1) was kept constant and the other protein (OBF-1 or POU domain) was titrated at increasing concentrations (Fig. 2C and D). As shown in Fig. 2C and D, the previously observed effects, for example the disruption of ternary complex formation in the presence of the αp6 phosphothioate modification, were entirely retained under both sets of conditions, indicating that these effects are not concentration dependent. In addition, in EMSAs done without OBF-1 it was clear that the various probes were equally well bound by the POU domain (data not shown) (see also Fig. 2A, C, and D). These data thus indicate that OBF-1 indeed interacts with several positions in the octamer sequence, both in the major groove, in particular on position +5 and as shown recently (3) on position +6, and on the backbone, as shown here. Furthermore, by using bromodeoxyuridine-substituted oligonucleotides, we could show that OBF-1 can be UV cross-linked to the cognate sequence, both in a ternary complex with the POU domain and also weakly by itself (data not shown).

OBF-1 interacts with the POU domain through residues both in the POU_S and the POU_H.

To identify which part(s) of the POU domain is necessary for selective interaction with OBF-1, we used Oct-1/Pit-1 POU domain chimeras (2) and assayed by EMSA their capacity to allow ternary complex formation (for details, see the legend for Fig. 3). This approach was chosen because the Pit-1 POU domain binds efficiently to the octamer motif, as a monomer like POU1, but does not interact with OBF-1 (Fig. 3, lane 3). When only the POU domain linker region was exchanged in POU1 the ternary complex formation was unaffected (lane 6). However, when the POU_S domain was swapped the ternary complex formation was either abolished completely (lane 4) or greatly reduced when only the most N-terminal part of the POU_S was exchanged (POU_SA subdomain; lane 8). By contrast, exchange of the POU_H did not completely abolish complex formation, although it significantly reduced it (lane 7). Together, these data indicate that both the POU_S and the POU_H are important for complex formation with OBF-1 and that the major contribution appears to come from the POU_S, in particular from the POU_SA subdomain. Recently, similar experiments that are in good agreement with our results were reported (3).

FIG. 3.

Formation of a complex among Oct-1/Pit-1 chimeras, OBF-1, and octamer site DNA is influenced by the origin of both the POU_S and the POU_H. In vitro-translated proteins were assayed by EMSA for their ability to interact by using an octamer-containing DNA probe. The four-letter nomenclature (2) beside the schematic representations of the POU domains reflects the structures of the chimeras: the first two capital letters indicate the origin of the POU_SA and the POU_SB, respectively, the third, lowercase letter symbolizes the POU linker, and the fourth, uppercase letter denotes the POU_H domain. O or o, Oct-1; P or p, Pit-1. All reaction mixtures contained equal amounts of OBF-1 protein (lanes 2 through 9) or of POU domain (lanes 1 through 9). For the exact sequences, see Fig. 7.

Two residues at the very beginning of the first helix of the Oct-1 POU_S are crucial for interaction with OBF-1 in vitro and in vivo.

To identify the residues in the POU_S domain that are critical for ternary complex formation, we substituted amino acids at several positions with alanine residues. We chose residues that differ in Oct-1 and Pit-1 and are solvent accessible based on the Oct-1 POU-DNA cocrystal (16) (see Fig. 7 and 9). As shown in Fig. 4A, all substitutions tested (D5A, T15A, G28A, M34A, K36A, D41A, N54A, N72A, and D73A) left the ternary complex formation unchanged with the exceptions of L6A and E7A (lanes 3 and 4), both of which reside in the beginning of the first helix of the POU_SA subdomain. The double substitution L6A-E7A totally abolished OBF-1 binding activity in vitro (data not shown) (see below). To confirm the importance of these two residues in mediating interaction between OBF-1 and the POU domain, we set up an in vivo assay. For this we used Drosophila SL2 cells that, based on EMSAs or transfection experiments, do not contain any Oct or OBF-1 protein (not shown) (see Fig. 4C) and therefore allow us to test interactions in vivo without interference from endogenous proteins. In initial experiments we found that, in this system, transactivation of an octamer-containing reporter plasmid required the simultaneous expression of Oct factors (e.g., Oct-1 or Oct-2) and OBF-1, suggesting that the activation domains of Oct-1 and Oct-2 that had been previously identified through experiments done in mammalian cells (7, 22, 35, 36) are inactive in these Drosophila cells. In agreement with this, we found that the isolated POU domain was able to mediate activation of the reporter plasmid essentially as efficiently as the complete Oct-1 or Oct-2, provided OBF-1 was present (data not shown) (see below). As shown in Fig. 4B, the wt Oct-1 POU domain mediated a strong transactivation of the reporter plasmid by OBF-1; mutation of either position 6 or position 7 (POU1-L6A or POU1-E7A) strongly reduced activation, and substitution of both residues simultaneously completely eliminated in vivo recruitment of OBF-1 by the POU domain. As shown in the inset on the right of Fig. 4B, the different POU domains were expressed equally well.

FIG. 7.

Alignment of the POU domains used in this study. The Oct-1 POU domain sequence was chosen as a reference to which the other sequences were compared. Identical positions are indicated with a dot, and at divergent positions the amino acid is indicated. Above the sequences the borders of the regions A and B and the α helices (I through IV) are indicated, as well as the numbering of the amino acids. The asterisk above the sequence for the POU1 linker shows the position at which the POU domain was separated into two halves for the experiment presented in Fig. 8. At the bottom the sequence of Oct-1 is displayed again with the residues examined in this study and critical for interaction with OBF-1 indicated by solid arrows; residues that were mutated and did not interfere with OBF-1 binding are indicated by open arrows. Squares and circles indicate POU domain residues identified in other studies (3, 8) that, when mutated, disrupted (solid symbols) or did not disrupt (open symbols) association with OBF-1.

FIG. 9.

Model of the interaction surface between the POU domain/DNA complex and OBF-1 highlighting the positions identified in this study. The modeling program Insight II was used to generate this picture by using the coordinates of the Oct-1 POU domain/octamer motif cocrystal (16). Panels A and C represent 90° counterclockwise or clockwise rotations along the vertical axis of the projection presented in projection B. The octamer sequence DNA ATGCAAT is displayed in magenta. Backbone positions enhancing or eliminating association with OBF-1 when modified with a phosphothioate group are represented by green (p4, p5, p7, p8 and p9) or red (p6) balls, respectively. Blue balls represent the backbone position p3 that was found not to influence interaction with OBF-1. The position of the methyl group on G at position +3 that interferes with OBF-1 binding is indicated by a red ball to which an arrow points. The two central base pairs at positions +5 and +6 important for OBF-1 binding are represented thicker. The POU domain is displayed in white. Residues whose substitution by alanine eliminated interaction with OBF-1 are indicated in red, and residues whose substitution by alanine did not interfere with OBF-1 association are depicted in blue. The POU linker that connects both domains together is not visible in the crystal, but should be imagined to pass on the left of the DNA in projection A or in the back of projection B.

FIG. 4.

Two residues at the very beginning of the POU_S domain are essential for interaction with OBF-1 in vitro and in vivo. (A) In vitro-translated proteins with alanine substitutions in the POU specific moiety of the Oct-1 POU domain were assayed for interaction with OBF-1 by EMSA as indicated in Fig. 3. The scheme below the autoradiograph indicates the first and last amino acids of the two POU specific domains. (B) Drosophila SL2 cells were cotransfected with an octamer-containing luciferase reporter plasmid together with an Oct-1 POU domain expression vector in the absence (white bars) or in the presence (hatched bars) of an OBF-1 expression vector. On the left the relative luminescence (rlu) obtained with equal protein amounts is indicated for one representative experiment among several that gave similar results. Expression of the different POU domains was tested by EMSA with cellular extracts from the transfected cells and found to be identical, as shown in the inset presented on the right. Only the part of the gel corresponding to the POU/DNA complex is shown. + or − for OBF-1 indicates whether the extract originates from a transfection containing or lacking OBF-1, respectively. (C) In vitro-translated proteins with exchanges of the Pit-1 residues for their Oct-1 counterparts in the context of the chimeras OOoO, POoO, and PPoO were assayed for OBF-1 binding activity by EMSA. The chimera used and the substitution introduced are indicated to the right of each lane.

To further confirm the importance of residues 6 and 7 of the POU_S, we subsequently also exchanged them with their Oct-1 counterparts individually or in combination in the two chimeras POoO and PPoO and tested complex formation in vitro by EMSA. As shown in Fig. 4C, in either context the protein with the substitution R7E showed a larger gain of ternary complex formation than that with I6L, and the double substitution I6L-R7E showed a synergistic effect in both chimeras (Fig. 4C, lanes 7 and 9). In addition, these results indicate that the effect is not dependent on the DNA binding helix of the POU_S, since it is observed with both chimeras POoO and PPoO (DNA binding helix underlined). Together, these results all support the initial finding that the POU_SA subdomain contains key determinants for the selectivity of interaction with OBF-1, but that additional residues in the POU_H are also required (see below).

Two residues in the third, DNA recognition helix of the Oct-1 POU homeodomain interact specifically with OBF-1.

From the analysis with the Oct-1/Pit-1 chimeras it is apparent that the chimera OOoP shows a much-reduced affinity for OBF-1 when compared to OOoO, indicating that the homeodomain also plays an important role for interaction with OBF-1 (Fig. 3). To identify the critical position(s) we replaced with alanine the residues of the Oct-1 POU homeodomain that differ in Oct-1 and Pit-1, are solvent accessible, and lie roughly on the same face of the POU1/DNA cocrystal relative to the already identified positions in the POU_S and on the DNA (see also Fig. 9). The substitutions in the first candidate region we looked at (S107, E109, T110, N111, R113, and V114) did not produce any effect (data not shown) (see Fig. 7 and 9). By contrast, substitutions at two positions in the second candidate region produced a distinct effect (I159A) or a weak effect (K155A; Fig. 5, lanes 3 and 2), while substitutions at two other positions had no effect (N160A and P161A; Fig. 5, lanes 4 and 5). When tested in the absence of OBF-1 all these mutant POU domains were found to bind equally well to the probe (data not shown), indicating that the reduction in interaction with OBF-1 did not reflect a reduced DNA binding capacity. To further analyze this, the Oct-1 counterparts of these different residues were introduced individually or in combinations into the OOoP chimera, which interacted only weakly with OBF-1 (Fig. 3). When the substitution R155K was present (Fig. 5, lanes 7, 11, and 13) a clear gain of ternary complex formation was always detected. These data suggest that of the positions tested in the POU_H, positions 155 and 159 are the most important for interaction with OBF-1, although their relative context appears to play a role (OOoO versus OOoP). Similar conclusions were also obtained in a very recent study, where it was shown that the POU_H positions K155 and I159 as well as the conserved positions N151 and R158 are critical for interaction with OBF-1 (3).

FIG. 5.

Residues at the end of the POU_H are critical for interaction with OBF-1. In vitro-translated proteins with alanine substitutions in the POU homeodomain of Oct-1 were assayed for OBF-1 binding activity by EMSA (lanes 1 through 5). In the Oct-1/Pit-1 POU domain chimera OOoP amino acids were exchanged individually (lanes 7 to 10) or in combination (lanes 11 to 13) for their Oct-1 counterparts. The substitution(s) introduced and its position(s) are indicated at the right of the figure. In the case of multiple substitutions only the introduced amino acids are indicated (lanes 11 to 13), and their positions correspond to those in lanes 7 to 10. Lane 14 contains only the probe.

Other members of the POU family can interact with OBF-1.

Based on the above results we made predictions about the minimal requirements for a POU domain to interact with OBF-1. The POU domain should be bound to a permissive DNA binding site (e.g., A at positions +5 and +6), and there should be either a leucine at position 6 or a glutamate at position 7 of the POU_S or both, depending on the context (Oct-1 versus Pit-1 POU_S surface composition); in addition, in the POU_H positions 155 and 159 also play a critical role depending on the context (Oct-1 versus Pit-1 POU_H surface composition). We looked for POU family members fulfilling all or part of these criteria and first tested the zebra fish pou-2 (10) POU domain. The POU domain of zebra fish pou-2 has positions E7 and K155 conserved and a divergent position T6 compared with Oct-1 (see Fig. 7). However, this POU domain showed OBF-1 binding activity comparable to that of the Oct-1 POU domain (Fig. 6A, lane 7). Thus, it seems that the negative charge of E7 is sufficient to allow interaction with OBF-1 and that in this context the divergent position T6 is not critical.

FIG. 6.

Interaction between OBF-1 and the POU domain of other POU family members. (A) In vitro-translated proteins assayed for OBF-1 binding activity by EMSA. Oct-1 POU domain (lanes 1 and 2), Oct-6 POU domain or derivatives thereof (lanes 3 to 6), and zebra fish pou-2 (zfpou2) POU domain (lanes 7 and 8). For Oct-6 POU the amino acids substituted with their Oct-1 counterparts are indicated to the right of each lane (lanes 4 to 6). (B) Transient-transfection assay in Drosophila SL2 cells. A luciferase reporter under the control of wt (solid bars) or mutated (open bars) octamer sites was introduced into SL2 cells together with an empty expression vector (lanes 1 and 2) or a vector expressing Oct-1 (lanes 3 and 4), Oct-6 (lanes 5 and 6) or Oct-6-S6L/D7E (lanes 7 and 8). Even-numbered lanes also contained an expression vector for OBF-1. The total amounts of expression vector in the different samples were equalized by the addition of empty vector. Forty-eight hours after DNA addition an aliquot of each sample was processed for luciferase assays (upper part panel B) and the rest was used for preparation of a whole-cell extract followed by EMSA with an octamer site probe to control for protein expression (lower part of panel B; only the part of the gel comprising the Oct protein/DNA complex is shown).

We have previously shown that neither the Oct-4 nor the Oct-6 POU domain interacts with OBF-1 in vitro (34). In the case of Oct-6, while position K155 in the POU_H is conserved, positions 6 and 7 in the POU_S domain are divergent from the corresponding positions in Oct-1 (Fig. 7). We therefore substituted these two divergent positions with their Oct-1 counterparts individually or in combination and tested the capacity of the resulting proteins to interact with OBF-1 in vitro. As shown in Fig. 6A, POU6-S6L interacted efficiently with OBF-1, while POU6-D7E did not (lanes 3 to 5); yet, the double substitution POU6-S6L-D7E showed a synergistic effect and allowed very efficient ternary complex formation (lane 6). To substantiate these findings in vivo, we introduced the above-mentioned double amino acid changes into full-length Oct-6 to create Oct-6-S6L-D7E and tested whether this protein was responsive to OBF-1 in a transient-transfection assay done in SL2 cells. For these experiments we cotransfected a reporter plasmid under the control of multiple octamer sites (wt or mutated) as well as expression vectors for Oct-1, Oct-6 wt, or Oct-6-S6L-D7E, with or without an OBF-1 expression vector. As described above, on their own none of the tested Oct factors activated the reporter plasmid (Fig. 6B), in spite of the fact that they were efficiently expressed in transfected cells, as shown at the bottom of Fig. 6B. However, when OBF-1 was cotransfected with Oct-1 the reporter was activated approximately 14-fold, and when it was cotransfected with Oct-6-S6L-D7E the reporter was activated about 15-fold. In contrast to what was anticipated based on the in vitro results, in SL2 cells cotransfection of Oct-6 wt together with OBF-1 also activated the reporter, but only ca. sixfold. Although no interaction had been detected with OBF-1 in vitro, either with the isolated POU domain (Fig. 6A) or with the full-length Oct-6 protein (data not shown), it is conceivable and supported by our data that Oct-6 may have a weak potential for interaction which, however, was not detected under stringent EMSA conditions. However, it is clear that changing the amino acids at positions 6 and 7 in the Oct-6 POUS domain to their Oct-1 counterparts results in a protein that in vivo activates in the presence of OBF-1 better than the wt protein, most likely reflecting enhanced recruitment of this coactivator.

In vivo OBF-1 can clamp artificially separated moieties of the POU domain on the DNA and thereby activate a reporter gene.

The data presented in this paper indicate that OBF-1 interacts with the POU domain by making selective contacts with specific residues in both the POU_S and the POU_H and also by touching several positions in the major groove and on the backbone of the octamer site. Thereby, OBF-1 appears to close the protein ring around the DNA, as the POU domain itself already surrounds the DNA on three sides: from the top (POU_S), from the bottom (POU_H), and from the back (POU linker; see Fig. 9). We therefore wondered whether OBF-1 might be able to tether to the DNA a POU domain that has been artificially separated into two halves in the middle of the linker. For this we expressed in SL2 cells the POU_S and/or the POU_H separately, in the presence or in the absence of OBF-1, and measured transactivation of the reporter plasmid. As shown in Fig. 8, when either of the POU domain moieties was expressed by itself, inclusion of OBF-1 in the transfection had no effect and no transactivation was observed. By contrast, when both the POU_S and the POU_H were coexpressed—from separate plasmids—addition of OBF-1 led to strong transactivation of the reporter plasmid, to a level at least equal to that obtained with the intact POU domain. These data strikingly demonstrate that for the formation of a ternary complex on DNA OBF-1 requires both moieties of the POU domain and further suggest that OBF-1 may have the capacity to tether the POU domain on the DNA, perhaps stabilizing it.

FIG. 8.

In vivo OBF-1 can recruit artificially separated halves of the POU domain to the DNA. (A) Drosophila SL2 cells were transiently transfected with an octamer-containing luciferase reporter plasmid and expression vectors encoding the entire Oct-1 POU domain, or its individual domains POU_S and POU_H, in the presence (hatched bars) or in the absence (open bars) of an OBF-1 expression vector, as indicated. Solid bars represent transfections performed with the same expression vectors but with a reporter plasmid containing mutated octamer sites. The figure shows the relative luminescence (rlu) obtained in a representative experiment among several that gave similar results. (B) Schematic representation of the POU domain derivatives used for these experiments.

DISCUSSION

In this study we have investigated the requirements for a POU domain to selectively interact with OBF-1 by analyzing parameters in the DNA and in the protein.

We assayed the footprint of OBF-1 in the ternary complex by a methylation interference assay and found that methylation of the guanine at position +3 of the octamer site (ATGCAAAT) interferes with OBF-1 binding. Furthermore, we modified the DNA backbone with phosphothioate groups and analyzed the effect on ternary complex formation. This procedure identified several positions on both DNA strands that influence interaction with OBF-1 when modified. In the Oct-1 POU domain, using amino acid substitutions, we identified several residues in the POU_S (L6 and E7) and the POU_H (K155 and I159) that are crucial for interaction with OBF-1 in vitro and also in vivo. Thus, in contrast to the other POU-domain-interacting proteins like VP16 and SNAPc/PTF, OBF-1 selectively interacts with both the POU_S and the POU_H as well as with the DNA. Strikingly, we found that in vivo OBF-1 is capable of recruiting separated POU domains to the promoter DNA, thereby leading to transcriptional activation of the reporter gene. This genetically demonstrates that OBF-1 needs to interact with both moieties of the POU domain and supports the notion that this coactivator may stabilize the POU domain on the DNA.

Ternary complex formation is influenced by a number of specific and nonspecific DNA determinants.

So far for formation of a ternary complex no sequence requirements have been identified outside the octamer motif for OBF-1. However, within the octamer motif multiple positions are important. The base pair at position +5 (ATGCAAAT) has been identified earlier as being essential for interaction with OBF-1 (4, 8), and recently position +6 and, to a lesser extent, positions +3 and +4 have also been found to be important (3). Here we showed by a different technique, methylation interference, that the guanine residue at position +3 is critical for the formation of a complex containing OBF-1.

Furthermore, by introducing phosphothioate modifications at various positions in the DNA backbone, we could show that several of these positions also modulate ternary complex formation, indicating that they are in close proximity of OBF-1. In particular, backbone position p6, when modified with a phosphothioate, completely disrupted interaction with OBF-1 without affecting the interaction with the POU domain. Interestingly, the identified backbone positions lay spatially between the critical residues identified in the POU_S (positions 6 and 7) and the POU_H (positions 155 and 159) (see below and Fig. 9). In addition, these backbone positions flank the major groove where the reactive groups of the identified bases at positions +3, +4, +5, and +6 are located. These data suggest that OBF-1 lies over this region in the major groove and makes interactions with several bases as well as with the backbone crests. The hydroxyl radical footprints performed by Verrijzer et al. (37, 39) examining the interaction between the POU domain and the DNA showed an unprotected deoxyribose window between positions +2 and +4 on the sense strand (ATGCAAAT) and at position +8 on the antisense strand (TACGTTTA). This window seems to be partially filled by OBF-1, since phosphothioates at positions +p4 and +p5 on the sense strand (ATGpCpAAAT) and positions αp7, αp8, and αp9 on the antisense strand (TACGTTTpApTp) increase ternary complex formation. It had been shown earlier that at a very high concentration the N-terminal amino acids 1 to 118 of OBF-1 can interact with the octamer site even in the absence of POU domain (4), and in line with this, we were able to cross-link with UV OBF-1 to a bromodeoxyuridine-substituted octamer probe (data not shown). Recently, by a DNase I footprint assay done with an artificial probe consisting of an Ig(κ) octamer close to an H2B octamer, it was shown that OBF-1 can stabilize the POU domain on the DNA (3), albeit weakly. Yet, in earlier experiments done by EMSA on single-site probes, OBF-1 was found neither to increase the on rate of POU domain binding to the DNA nor to decrease its off rate (34).

Structure of the POU domain on the DNA.

For our analysis we assumed that the POU domains of Oct-1 and Pit-1 as well as the chimeras thereof bind as a monomer in a roughly similar conformation to the octamer-site probe used for the EMSAs. In the cocrystal of POU1 bound to DNA (octamer site), the POU domain binds as a monomer interacting with opposite faces of the DNA (16); by contrast, in the recently described cocrystal of the Pit-1 POU domain bound to DNA (prolactin gene promoter-derived site), the POU domain binds as a dimer with each POU domain binding to one face of the DNA only (13), thus revealing surprisingly different overall arrangements of the POU domain in the two crystals. Yet, the conformations of the individual POU domain subunits (POU_S and POU_H) regarded on their own appear to be very similar in the two cocrystals. Clearly, the nature of the DNA site appears to be of crucial importance in defining the overall spatial arrangement of a POU domain bound to DNA. When we compared the Pit-1 POU domain and the chimera PPoO (containing only the POU_S domain from Pit-1) on different probes, we found that these proteins indeed bind as a dimer on a prolactin promoter probe but as a monomer on an octamer probe; on the other hand, POU1 bound as a monomer on either of these probes (data not shown). Thus, on the basis of these considerations it can be assumed that Pit-1 POU binds to the octamer site in a conformation similar to that of POU1; this is further supported by the finding that the chimera PPoO/6L7E recruits OBF-1 efficiently (Fig. 4C).

Multiple residues in the POU_S and the POU_H mediate interaction with OBF-1.

Using Pit-1/Oct-1 POU domain chimeras, we found that the POU_S and the POU_H both contribute to association with OBF-1, and other recent experiments are in agreement with this (3). On the basis of these results we tested by mutation amino acids that were different in Pit-1 and Oct-1 and that are deemed to be located on the surface of the POU domain based on the available crystallographic data. We found that mutation with change to alanine at position L6 or E7 in the Oct-1 POU_S was sufficient to severely disrupt in vitro association with OBF-1 (Fig. 3) and that mutation of both residues completely abolished the interaction. In addition, these findings could be entirely recapitulated in an in vivo transactivation assay in which wt or mutant POU1 domains were used to recruit OBF-1 to the promoter of a reporter gene. Finally, introduction of the Oct-1 amino acids at the corresponding positions in the POoO or PPoO chimeras resulted by and large in the expected increase in association with OBF-1.

In addition to positions 6 and 7, which are divergent among different POU domains, the conserved positions E9, L10, R49, L53, L55, S56, N59, and M60 have recently been found to be critical for association between the POU1 domain and OBF-1 (3). By contrast, several other conserved residues (e.g., Q11, K17, F57, K58, or K64) (3) as well as nonconserved residues (e.g., D5, E8, K14, T15, D29, M34, K36, or D41, examined here) can be mutated with change to alanine without deleterious effect on the interaction with OBF-1 (summarized in Fig. 7; see also Fig. 9). Thus, it appears that in the POU_S OBF-1 requires a number of residues for interaction but discriminates between a permissive (e.g., Oct-1) and a nonpermissive (e.g., Pit-1) POU domain on the basis of only two residues at the very beginning of the first helix.

A somewhat similar situation was found for the selective interactions between OBF-1 and the POU_H. Results from this study as well as data from the report by Babb et al. (3) demonstrated that replacement with alanine of several conserved (e.g., E118, E130, or R152) or nonconserved amino acids (e.g., T107, E109, T110, N111, R113, V114, N160, or P161) does not disrupt association with OBF-1. By contrast, a few residues at the end of the third POU_H α-helix, in particular K155 and I159, and also the conserved N151 and R158 (3) are essential, as shown above. In the context of the chimera OOoP we found only position 155 to produce a significant increase in association with OBF-1 when substituted with its Oct-1 counterpart (R155K) (see Fig. 5, 7, and 9).

A reason that could account for this effect is the unraveled tertiary structure found in the third helix of the POU1_H and POU2_H (16, 33). Even though conserved at the level of primary structure, the last four residues of this helix in the two proteins show a tertiary structure divergence. The same unraveled tertiary structure in the C-terminal region of the third helix is also found in the Pit-1 POU_H domain (in addition to a divergence in primary structure) (13).

Given these findings, it appears that most, if not all, of the selectivity of interaction with OBF-1 rests with the few POU domain residues that were identified here: L6, E7, K155, and I159. This was further confirmed by showing that zebra fish pou-2, which fulfills most of the above-mentioned criteria, does indeed associate efficiently with OBF-1. In addition, we demonstrated that Oct-6 can interact well with OBF-1 in vitro, provided positions 6 and 7 are converted to their Oct-1 counterparts.

Requirement for VP16 and SNAPc to interact with the POU domain.

Specific interaction of the POU domain on the octamer with the two transcriptional coactivators VP16 and SNAPc has already been precisely described. The herpes simplex virus protein VP16 has been shown to discriminate Oct-1 from Oct-2 through position 22 in the POU_H domain (Glu for POU1 versus Ala for POU2). In addition, for efficient complex formation VP16 requires a conserved DNA-flanking region adjacent to the octamer motif as well as a host cell factor.

It has been shown recently that the RNA Pol II- and Pol III-specific coactivator SNAPc/PTF interacts with the POU1_S residue E7 (21) through the SNAP190 subunit (41). It is interesting that position E7 of the POU_S is the sole crucial binding determinant identified for interaction with SNAP190, and indeed SNAP190 shows a weak homology to the region in the N terminus of OBF-1 that interacts with the POU domain. Both VP16 and SNAP190 interact selectively with one residue of either the POU_H or the POU_S, respectively. By contrast, OBF-1 interacts specifically with a number of residues located in both POU domain moieties, and it was shown recently that OBF-1 and VP16 can bind simultaneously to the POU domain on the DNA (3).

OBF-1 can recruit artificially separated POU domain moieties to the DNA.

The results presented in this study suggest that OBF-1, contacting both moieties of the POU domain and closing the protein ring around the DNA, may in part have a function equivalent to that of the POU linker in holding the POU domain together and possibly stabilizing it on the octamer site. We therefore wondered whether OBF-1 might help to recruit to the octamer site two artificially separated POU domain moieties. To test this hypothesis an in vivo transactivation assay was used, and we could show that the reporter plasmid was indeed efficiently activated when both POU_S and POU_H as well as OBF-1 were present (Fig. 8). By contrast, when only the POU_S or the POU_H was present together with OBF-1, no activation was observed. This result strongly suggests that in this setting OBF-1 in some way replaces the POU linker and helps to recruit and stabilize the two POU domain moieties on the DNA. Since OBF-1 carries a strong transcription activation domain in its C-terminal part (8, 27), formation of the ternary complex then leads to activation of the reporter plasmid. Babb et al. (3) also recently came to the same conclusion on the basis of DNase I footprinting experiments as well as in vivo transactivation by using a VP16-POU1 fusion protein together with a truncated OBF-1 (amino acids 1 to 118) lacking the transcription activation domain. This VP16-POU1 fusion protein did not activate transcription by itself (surprisingly), but it did so in the presence of the truncated OBF-1, suggesting that in this case OBF-1 acted to stabilize the VP16-POU1 molecule on the DNA. We also found that a similar VP16-POU1 fusion protein is indeed inactive in vivo unless OBF-1 is present; in addition, we found that, unlike POU1, this VP16-POU1 protein does not bind to DNA in an EMSA, presumably due to some steric hindrance caused by VP16 (data not shown). One can thus speculate that in this case also OBF-1 in some way rearranges the POU domain such that DNA binding is rendered possible in spite of the negative effect of VP16.

Taken together, all these findings suggest that the formation of a ternary complex among the POU domain, OBF-1, and the DNA is a highly complex process in which each partner may influence the other two. Although the POU domain is clearly able to bind to DNA by itself and can be viewed as recruiting OBF-1, one can also speculate that OBF-1 in fact helps the POU domain to organize itself on the DNA and thereby stabilizes it. Such a coactivator not only may thus play a role at the level of transcription initiation but also could, for instance, help to form a stable transcription complex locked on the DNA and allowing multiple rounds of transcription to take place.